linux-next: build failure after merge of the rdma-leon tree

linux-next: build failure after merge of the rdma-leon tree

[Stephen Rothwell]

Hi Leon,

After merging the rdma-leon tree, today's linux-next build (powerpc
ppc64_defconfig) failed like this:

ERROR: ".rdma_addr_size" [drivers/infiniband/core/rdma_ucm.ko] undefined!
ERROR: ".rdma_addr_register_client" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_addr_size" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_addr_unregister_client" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_translate_ip" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_copy_addr" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_resolve_ip" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_addr_cancel" [drivers/infiniband/core/rdma_cm.ko] undefined!
ERROR: ".rdma_resolve_ip_route" [drivers/infiniband/core/ib_sa.ko] undefined!
ERROR: ".addr_init" [drivers/infiniband/core/ib_core.ko] undefined!
ERROR: ".rdma_addr_find_l2_eth_by_grh" [drivers/infiniband/core/ib_core.ko] undefined!
ERROR: ".addr_cleanup" [drivers/infiniband/core/ib_core.ko] undefined!

I don't know what has caused this, but the rdma-leon tree shares a lot
of patches (but not commits) with the rdma tree and so needs to be
cleaned up. Please sort out your work flow.

I have dropped the rdma-leon tree for today.

-- 
Cheers,
Stephen Rothwell

Re: linux-next: build failure after merge of the rdma-leon tree

[Stephen Rothwell]

Hi Leon,

On Fri, 13 May 2016 10:08:07 +1000 Stephen Rothwell <sfr@canb.auug.org.au> wrote:
>
> After merging the rdma-leon tree, today's linux-next build (powerpc
> ppc64_defconfig) failed like this:
> 
> ERROR: ".rdma_addr_size" [drivers/infiniband/core/rdma_ucm.ko] undefined!
> ERROR: ".rdma_addr_register_client" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_addr_size" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_addr_unregister_client" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_translate_ip" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_copy_addr" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_resolve_ip" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_addr_cancel" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_resolve_ip_route" [drivers/infiniband/core/ib_sa.ko] undefined!
> ERROR: ".addr_init" [drivers/infiniband/core/ib_core.ko] undefined!
> ERROR: ".rdma_addr_find_l2_eth_by_grh" [drivers/infiniband/core/ib_core.ko] undefined!
> ERROR: ".addr_cleanup" [drivers/infiniband/core/ib_core.ko] undefined!
> 
> I don't know what has caused this, but the rdma-leon tree shares a lot
> of patches (but not commits) with the rdma tree and so needs to be
> cleaned up. Please sort out your work flow.
> 
> I have dropped the rdma-leon tree for today.

Note that the following merge of the rdma-leon-testing tree compiled
fine and there is only one commit in there that is not in the rdma-leon
tree ... (though there are also several merge commits).

-- 
Cheers,
Stephen Rothwell

Re: linux-next: build failure after merge of the rdma-leon tree

[Doug Ledford]

[-- Attachment #1: Type: text/plain, Size: 2328 bytes --]

On 05/12/2016 08:08 PM, Stephen Rothwell wrote:
> Hi Leon,
> 
> After merging the rdma-leon tree, today's linux-next build (powerpc
> ppc64_defconfig) failed like this:
> 
> ERROR: ".rdma_addr_size" [drivers/infiniband/core/rdma_ucm.ko] undefined!
> ERROR: ".rdma_addr_register_client" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_addr_size" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_addr_unregister_client" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_translate_ip" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_copy_addr" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_resolve_ip" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_addr_cancel" [drivers/infiniband/core/rdma_cm.ko] undefined!
> ERROR: ".rdma_resolve_ip_route" [drivers/infiniband/core/ib_sa.ko] undefined!
> ERROR: ".addr_init" [drivers/infiniband/core/ib_core.ko] undefined!
> ERROR: ".rdma_addr_find_l2_eth_by_grh" [drivers/infiniband/core/ib_core.ko] undefined!
> ERROR: ".addr_cleanup" [drivers/infiniband/core/ib_core.ko] undefined!
> 
> I don't know what has caused this,

This was, I think, a botched merge between a fix in Leon's
topic/fix-core and topic/rdma-rw-api, but I could be wrong.

> but the rdma-leon tree shares a lot
> of patches (but not commits) with the rdma tree and so needs to be
> cleaned up. Please sort out your work flow.

This is an entirely separate issue.  Leon based all of his topic
branches on Linus' 4.6-rc7.  Even though I already have my initial
for-next tag available (it's based on 4.6-rc6 + my k.o/for-4.6-rc branch
to pick up a few commits that were after 4.6-rc6) and it has a number of
already commited for 4.7 patches on it.  He then committed a bunch of
those same patches in his topic branches and then merged them together
to make the branch you pulled.  That is an entirely broken workflow.  He
must base his topic branches, and his final branch, on my for-next tag
and that way exclude the patches I've already taken from the ones he is
testing.  I'm not sure he's aware of that, but as of this email, he
certainly is now.

> I have dropped the rdma-leon tree for today.
> 


-- 
Doug Ledford <dledford@redhat.com>
              GPG KeyID: 0E572FDD



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

linux-next: build failure after merge of the rdma-leon tree

[Stephen Rothwell]

Hi Leon,

After merging the rdma-leon tree, today's linux-next build (x86_64
allmodconfig build) failed like this:

In file included from include/linux/swab.h:4:0,
                 from include/uapi/linux/byteorder/little_endian.h:12,
                 from include/linux/byteorder/little_endian.h:4,
                 from arch/x86/include/uapi/asm/byteorder.h:4,
                 from include/asm-generic/bitops/le.h:5,
                 from arch/x86/include/asm/bitops.h:517,
                 from include/linux/bitops.h:36,
                 from include/linux/kernel.h:10,
                 from include/linux/list.h:8,
                 from include/linux/module.h:9,
                 from drivers/infiniband/hw/mlx5/qp.c:33:
drivers/infiniband/hw/mlx5/qp.c: In function 'mlx5_ib_modify_wq':
drivers/infiniband/hw/mlx5/qp.c:4910:48: error: 'struct mlx5_ib_port' has no member named 'q_cnt_id'
    MLX5_SET(rqc, rqc, counter_set_id, dev->port->q_cnt_id);
                                                ^
include/uapi/linux/swab.h:113:54: note: in definition of macro '__swab32'
 #define __swab32(x) (__u32)__builtin_bswap32((__u32)(x))
                                                      ^
include/linux/byteorder/generic.h:93:21: note: in expansion of macro '__cpu_to_be32'
 #define cpu_to_be32 __cpu_to_be32  
                     ^
drivers/infiniband/hw/mlx5/qp.c:4910:4: note: in expansion of macro 'MLX5_SET'
    MLX5_SET(rqc, rqc, counter_set_id, dev->port->q_cnt_id);
    ^

Caused by commit

  cbd0624cc927 ("IB/mlx5: Expose Q counters groups only when they are supported by FW")

interacting with commit

  fbcb2b5c3749 ("B/mlx5: Add port counter support for Receive WQs")

This needed to be fixed up in the merge commit

  3064f9066da1 ("Merge branch 'topic/q_counters' into rdma-next")

I have dropped the rdma-leon tree again for today.

-- 
Cheers,
Stephen Rothwell